Volume 39, Issue 1 Transparency in repeated procurement: when hiding is better Berardino Cesi University of Rome Tor Vergata
Adriano Di Natale University of Rome Tor Vergata
Abstract In this paper we study the effect of transparency on the willingness to collude in repeated procurement competitive tenderings. We allow the buyer to postpone the revelation of the winner's identity and show that such a policy may make collusive agreements less stable in both simultaneous and sequential competitive procedures. When the buyer postpones the revelation of the winner's identity in a scenario in which colluding and "honest" (never colluding) firms participate to the same tendering, the threat of a retaliation, by means of an aggressive bidding by the colluding firms, is weakened and collusion is less stable
Citation: Berardino Cesi and Adriano Di Natale, (2019) ''Transparency in repeated procurement: when hiding is better'', Economics Bulletin, Volume 39, Issue 1, pages 14-23 Contact: Berardino Cesi -
[email protected], Adriano Di Natale -
[email protected]. Submitted: October 23, 2018. Published: January 10, 2019.
1
Introduction
Most of the procurement laws and guidelines around the word (UNICITRAL Model Law on Public Procurement, EU and Russian legislation) con…rm an increase in the level of transparency requirements as the most e¤ective way to …ght and eradicate corruption in public procurement (Bergot et al. 2010). Information about key stages of procurement processes, decisions and outcomes are generally publicly available to numerous public and concerned parties, because of mandatory law requirements. However, economic literature has also pointed out that transparency requirements may expose public authorities to the risk of collusive behaviors among public procurement stakeholders. Morozov and Podkolzina (2013) show how transparency policies in procurement may induce …rms to agree on collusive o¤ers and enjoy extra pro…t from the collusive behavior. In line with the extant literature, those transparency requirements may: i) provide clear focal points, ii) reduce the cost of monitoring other members of a collusive agreement iii) enforce a more e¤ective retaliation against cartel’s deviators. Albano et al. (2006) suggest precautions that might be adopted to tackle this issue: publish anonymous bids, limit the number of bids published or the bids in their last digits. The impact of public information dissemination on bidding behavior in procurement is also con…rmed in laboratory experiments and econometric studies (See Kagel et al. 1987; Goeree, O¤erman, (2002). For the empirical analysis see De Silva et al. 2003; Albano and Spagnolo (2005); De Silva et al. 2007). In this paper we show that ex post information disclosure of the identity of the winner may a¤ect the stability of collusion according to whether the buyer runs simultaneous or sequential auctions. Albano and Spagnolo (2010) provide a theoretical model to study collusion under both sequential and simultaneous auctions when the buyer splits the contract in two lots. To tackle the e¤ect of transparency on collusion we modify their model by introducing two new ingredients: 1) a honest …rm never involved in a cartel that may decide either to bid aggressively or to accommodate 2) the possibility for the buyer to postpone the announcement of the winner’s identity. We show that when there exists a …rm not involving in a cartel the sustainability of a collusive agreement always decreases in the lag the buyer waits to reveal the identity of the winner. In particular, this results holds in both simultaneous and sequential procedures, suggesting that less transparency, under some circumstances, is likely to …ght collusion in repeated procurement. In a sequential format, the buyer sequentially launches the auction for each contract within the same period, participants bid, …rst for one lot and once the auction for the …rst lot is over they bid for the second. At the end of the …rst round of bids, the buyer awards the lot and
2
it proceeds tendering the second contract. The identity of the winner will be kept secret for T periods and it will be disclosed at the end of each auction in each lot once these T periods are over; in other words at the end of the …rst auction, occurring after the T periods, the identity of the previous winner has made public. In the simultaneous format instead the buyer simultaneously launches the auctions for both contracts, participants bid for both contracts and the two lots are awarded at the end of period. The buyer keeps the identity of the winner secret for T periods and it will be disclosed at the end of the auction taking place (simultaneously on both lots) after these T periods are over. Given the presence of the honest …rm, in each period the colluding …rms have believes about the behavior of such a …rm that, with a certain probability, may either bid a reserve price or bid aggressively. The "soft" bidding of the reserve price by the honest …rm is in turns justi…ed by the idea that such a …rm, although honest, has believes about the "possible" collusive behavior of the other …rms and may try to gain an (expected) collusive pro…t even without taking part to a collusive scheme. In this scenario the colluding …rm cannot optimally start the punishment path because, without transparency, there exists the risk of punishing a …rm that is sticking to the collusive behavior.1 In line with the collusive behavior of Albano and Spagnolo (2010) we assume that in the sequential auction bidders apply a "bid rotation" strategy such that in each period the winning candidate of the collusive scheme bids the reserve price while the other uses a phony bid slightly higher than the reserve price. In a simultaneous auction instead the two …rms follow a "split the cake" collusive strategy such that the two lots are split in a way that the winning candidate bids the reserve price in each lot. The introduction of a honest …rm in the setting of Albano and Spagnolo (2010) …ts the real procurement scenario in which there exist some "honest" …rms that are not involved in a collusive agreement even thought collusion still exists. The existence of a honest …rm, that ex ante may steal the two lots, increases the willingness to deviate for a colluding …rm because, if the buyer has not revealed the identity of the winner, an immediate retaliation cannot be optimal. Our main conclusion is that the principle of transparency in public procurement may con‡ict with the necessity to …ght collusion since, under transparency, any possible deviators may be readily spotted and punished by the cartel. The public buyer can reduce the degree of transparency to make potential cartel unstable when also honest …rms are expected to participate to the competitive tenderings. In the section 2 we state the model and give the main result. Section 3 concludes. 1
This scenario is clearly similar to the framwerok of demand shocks in Green and Porter (1984) where
colluding …rms cannot fully observe whether the pro…t loss is due to a demand shock or a cheating behavior.
3
2
The model
We follow the set up in Albano and Spagnolo (2010). A public buyer wants to procure two procurement contracts (A and B) for an in…nite number of periods, each denoted by t, with t = 0; :::; 1, through a lowest-price sealed-bid auctions with a reserve price, either sequentially or simultaneously, when only …xed price contracts are available to the buyer. Three longlived …rms, denoted by subscripts i = 1; 2; 3, can deliver the project. All …rms share the same discount factor, , and production cost, c, for undertaking two equivalent procurement contracts. Each …rm perfectly know its own and the rival’s cost.2 The buyer commits to the same publicly announced reserve price, r, for each contract, with r > c. Let us to keep the same jargon in Albano and Spagnolo (2010) and de…ne the highest pro…t given by v = r
c. In Albano and Spagnolo (2010) all the relevant informations
are disclosed by the buyer (bids and identity of the …rm winning the competitive tendering), here instead we aim to study the e¤ect of ex post disclosure policies on the willingness to collude and we move one step ahead respect to the disclosure policy suggested in Albano and Spagnolo (2006).3 We propose an alternatively way of proceeding: the buyer is allowed to delay the public announcement of the winner’s identity, until one of the successive rounds of the repeated auction is completed and the participants submitted their bids. The game for the two formats are the following: Sequential format : At time t, the buyer sequentially launches the auction for each contract; participants bid, …rst for the …rst lot and once the auction for the …rst lot is over they bid for the second. At the end of each sub-round of bids, the buyer awards the lot to the lowest bid, then, it proceeds tendering the second contract. At the end of each sub-round it does not reveal the identity of the winners. The identity will be kept secret for T periods and it will be disclosed at the end of each auction for each lot at t + T . In other words, the buyer will announce the winner of the …rst lot (let’s say) A, launched at t, only after the conclusion of the bids for the lot A, at time t + T . Simultaneous format : At time t, the buyer simultaneously launches the auction for both 2
The assumption of complete information among participants has been commonly used in procurement
because it is realistic for many procurement situations (Albano and Spagnolo, 2010). An example is the consultancy procurement market in territory of Luxembourg: the market is settled down, with participants (the so-called big four) that are long-term rivals with massive know-how exchanges and knowledge of the market. The same assumption for models of repeated procurement is used in Albano et al. (2017) and Spagnolo and Calzolari (2009). 3 In particular, 1) Do not announce the winning bid price, 2) the winners’ identities,3) the losing bids or 4) the non-winners’ identities.
4
contracts; participants simultaneously bid for both contracts and, at the end of period, the buyer awards each lot to the lowest bid. At the end of each period it does not announce the winners’ identity of the lots, it will be kept secret for T periods and it will be disclosed at the end of the auctions taking place at t + T . In other words, the buyer will announce the winner of the lots, launched in t, only after the conclusion of the bids for the two lots at t + T . We follow Albano and Spagnolo (2010) in modelling a collusive scheme between …rm 1 and 2, but we also assume that …rm 3 is honest and it does not enter any collusive agreement and no communication is undertaken with …rm 1 and 2. The …rms 1 and 2’ strategies follows the ones in Albano and Spagnolo (2010). When colluding, …rm 1 and 2 cannot control for the bid of …rm 3 that can be either aggressive or not. On the other hand, …rm 3 does not know the existence of a possible cartel made by …rm 1 and 2, but it cannot exclude that it may occur. In any case it will never follow a collusive behavior because of its honesty. Thus ex ante …rm 1 and 2 in turn cannot exclude that, even not taking directly part to the cartel, …rm 3 could even bid the reserve price. To easily model this scenario we assume that …rm 1 and 2 have the following believes on the bidding behavior of …rm 3: …rm 3 bids aggressively (a price lower than the reserve price) with probability 1 whereas it bids the reserve price with probability . We assume that the tie-breaking rule is a ‡ip of the coin. The trigger-collusive strategies of each …rm depends on the format. In the sequential format …rms apply a "bid rotation" strategy, denoted by format they apply a "split-award" collusive strategy, R
S,
R,
whereas for the simultaneous
where:
: at time t, …rm i bids bi = r in both contracts at time t and bi > r at any time t + 1,
if up to time t …rm j 6= i has bid bj > r in both contracts; otherwise it reverts to the bid of the static equilibrium for ever. S
: at time t, …rm i bids bi = r in contract A and bi > r in contract B, if up to time t …rm
j 6= i has bid bj > r in contract A; otherwise it reverts to the bid of the static equilibrium for ever. When suppliers decide to collude they support the ring by the threat of reverting to competitive behavior forever in case a deviation is observed. This punishment is optimal in this symmetric environment as it minimaxes the deviator, exactly as in a symmetric Bertrand supergame. It is possible to note that the behavior of …rm 3, as "exogenously" modelled, does not a¤ect the standard static Nash equilibrium pro…ts of the auction. If …rm 3 bids the reserve price in a sequential format, the static collusive pro…t obtained by the winning candidate for the cartel is v, given the tie-breaking rule. In fact, with probability 5
…rm 3 bids the reserve price and the buyer ‡ips the coin to award the two lots between the winner from the cartel and …rm 3; the expected static pro…t for the winning member of the cartel is
1 v 2
+ 12 v =
v. Clearly the pro…t is zero with probability 1
argument the static collusive pro…t for the simultaneous format is
. By the same
v . 2
Because of the uncertainty about …rm 3’s behavior, …rm 1 and 2 start the punishment phase only when the identity of the winner is revealed because, in case one of the member of the cartel does not get the contract, it may be either because …rm 3 has won the auction or because of a deviation. Although we do not introduce in the action set of the …rms the possibility of revealing or nor their identity, the cheating …rm has clearly no incentive to make the result of the auction public. The result would be less straightforward for the honest …rm. In principle the honest …rm, if strategic, may …nd it pro…table to reveal or not its identity as long as the collusive …rms keep colluding and bidding high price. In this case, in fact, the honest …rm would increase its expected pro…t since there still exists the possibility to win by undercutting a collusive high price. However, as long as any cheating …rm may have no incentive to reveal their identity, it is not clear whether the revelation by the honest …rm would a¤ect the believes of the dishonest …rm about the possibility of being detected or detect a deviation. Clearly a revelation may protect a cheating …rm from retaliation but on the other hand a non revelation in turn does not necessarily induce a non-cheating …rm to start the punishment path. However in general the rules of the procedures …rms and buyer must stick to are assumed public knowledge and accepted by the participants when taking part to the procedure. These rules are enforceable by the buyer by means of the power of a court of law. However, it is important to stress that in practice the decision about the revelation does not stand in the …rms’ hands. United Nation General Terms and Conditions of Contract (Art.11 and 12) and the European Union procurement Directive 2014/24 explain that the buyer has the power to set speci…c disclosure policies in the tender documents (the General Terms and Conditions of contract) and it may oblige the suppliers to ask prior authorization before advertise or make public, for purposes of commercial/visibility advantage, any information and contractual relationship with the Contracting Authority itself. Since the request is scrutinized by the buyer, it is plausible that in our scenario the buyer denies such a publicity if it is willing to …ght possible cartels. If the winner does not stick to these transparency policy, i.e. by revealing some information whose revelation is forbidden, this would be a clear infringement of the law.4 4
We thank the referee for rising this point. This analysis could be formalized for future research in a richer
model in which players can also be strategic in setting the decision of revealing or not their identity.
6
Let’s de…ne the following expected pro…ts:5 Sequential format a) …rm 3 bids the reserve price and …rms i = 1; 2 collude, then the winning candidate of the cartel gains v, b) …rm 3 bids the reserve price and (one) …rm i = 1; 2 deviates, then the deviating …rm gains 2 v, c) …rm 3 bids aggressively, the pro…t for …rm i = 1; 2 is zero under but collusive and deviating behavior. Simultaneous format a) …rm 3 bids the reserve price and …rms i = 1; 2 collude, then the v , 2
winning candidate of the cartel gains
b) …rm 3 bids the reserve price and (one) …rm
i = 1; 2 deviates, then the deviating …rm gains 2 v, c) …rm 3 bids aggressively, the pro…t for …rm i = 1; 2 is zero under but collusive and deviating behavior The following Proposition gives the main result of the paper: Proposition 1 Let
R
and
S
be: T R
(1 +
R)
and
T R
2=0
1
The strategy pro…les
R
and
competitive tendering when when
2
S
3 T +1 S = 4 characterize a SNE respectively under a sequential R
and under a simultaneous competitive tendering
R.
Proposition 1 shows that less transparency about the identity of the winner (an increase in T ) would increase
S
and
R,
then it makes collusion less stable. This result is in line with the
standard argument about collusion in repeated game such that, the higher the transparency about the actions played by the players more e¤ective is the punishment enforcing the collusive agreement. When the actions (bids) of the …rms are made public, …rms are perfectly able to check each other after each auction, therefore the threat of retaliation (by means of an aggressive biding) is credible and it makes the collusive agreement stable. With lower transparency in fact the buyer makes such a retaliation less credible. Thus, we conclude that reducing transparency in repeated procurement reduces the risk of collusion in both simultaneous and sequential competitive procedures. See Figure 1 for a graphical explanation of this result for the case of sequential auctions (where we plot the critical discount factor for a "reasonable" lag in the revelation of the winner’s identity) 5
We assume that the deviation price of the cheating …rm is higher (because marginally close to the reserve
price) than the aggressive price of …rm 3.
7
3
Conclusions
Transparency has been considering one of the best practice against corruption in procurement. In this paper we tackle this argument by showing that a lower degree of transparency however is a good practice to …ght collusion when the buyer runs both simultaneously and sequential competitive procedures. We introduce the possibility for the buyer to postpone the winner’s identity in a repeated procurement scenario. Postponing such an information makes the retaliation strategy of colluding …rms more credible and increases the incentive for a maverick …rm to destroy a collusive agreement. Such a policy is more suitable when the market is characterized by the presence of "honest" …rms that would never enter a collusive agreement. This framework …ts some of real procurement scenarios where cartels and honest …rms coexist.
8
Appendix
Proof of Proposition 1 Consider T to the collusive path of
R
1. The collusive pro…t for each …rm when sticking
is: VC =
2 v 1 whereas when deviating at time t = 0, the discounted pro…ts are:
V
D
=
T 1 X
t
T
(2 v) +
v+
t=0
VD = R
t
(0)
t=T +1
with:
let’s de…ne
1 X
T
1 1
(2 v) +
T
v
as the lowest (critical) discount factor obtained from solving V C
V D = 0,
that is the net bene…t from collusion, then we have: VC
(1 + R )2 2 1 R
T R
VD = v
2
=0
(1)
Let’s take a simpli…ed version of the net bene…t: F( T R
The critical discount factor
R; T )
T R
=
(1 + )2
solving F (
R; T )
2
(2)
= 0 gives the lowest discount factor
such that collusion is stable. The collusive pro…t for each …rm, when sticking to the collusive path in
S,
is: 1
VC =
v 2
1
whereas when deviation at time t = 0, the discounted pro…ts are: V
D
=
T X
t
1 X
( 2v) +
t=0
t=T +1
that gives:
Where V C
T +1
1
VD =
2v
1
V D gives: 1
S
9
3 T +1 = 4
t
(0)
References Abreu, D J (1988) “On the Theory of In…nitely Repeated Games with Discounting” Econometrica 56, 383– 396. Albano, G L, B. Cesi and A. Iozzi (2017) “Public Procurement with Unveri…able Quality: the Case for Discriminatory Competitive Procedure” Journal of Public Economics 145, 14-26. Albano, G L and G. Spagnolo (2010) "Asymmetry and Collusion in Sequential Procurement: A "Large Lot Last" Policy" The B:E: Journal of Theoretical Econmics 10, 1-18. Albano, G L and G. Spagnolo (2005) "The Collusive Drawbacks of Sequential Auctions", Quaderni Consip http://www.consip.it/sites/consip.it/…les/896660_Quaderni_IX_2005.pdf Albano, G L, P. Buccirossi, G. Spagnolo and M.Zanza (2006) "Preventing Collusion in Procurement: A Primer" in HANDBOOK OF PROCUREMENT by N. Dimitri, G. Piga and G. Spagnolo, Eds., Cambridge University Press. Bertot, J C, P.T. Jaeger and J.M. Grimes (2010) "Using ICTs to create a culture of transparency: Egovernment and social media as openness and anti-corruption tools for societies" Government Information Quarterly 27, 264-271. Che, Y. K. (2008). “Procurement” in The New Palgrave Dictionary of Economics (2nd edition) by S. N. Durlauf and L. E. Blume (eds.) Da Silva, D G, T. Dunne, A. Kankanamge and G. Kosmopoulou (2007) "The impact of public information on bidding in highway procurement auctions" European Economic Review 52, 150–181 De Silva, D G, T. Dunne and G. Kosmopoulou (2003) "An empirical analysis of entrant and incumbent bidding in road construction auctions" Journal of Industrial Economics 51, 295–316.
EU Directive 2014/24 – EU Procurement Directive. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=celex%3A32014L0 European Commission (1996). Green Paper on “Public Procurement in the European Union: Exploring the Way Forward” adopted by the Commission on 27 November 1996, COM (96) 583. Goeree, J K and T. O¤erman (2002) "E¢ciency in Auctions with Private and Common Values: An Experimental Study" American Economic Review 92, 625-643. United Nations General Terms and Conditions of Contract. https://www.un.org/Depts/ptd/about-us/conditionscontract
10
Kagel, J H, L Harstad and D. Levin "Information Impact and Allocation Rules in Auctions With A¢liated Private Values: A Laboratory Study" Econometrica 55, 1275-1304 Mailath, G J and L. Samuelson (2006) Repeated Games and Reputations. Long Run Relationships, Oxford University: Oxford. Mailath, G J, V. Nocke and L. White (2008) "When the Punishment Must Fit the Crime: Remarks on the Failure of Simple Penal Codes in Extensive Form Games” Mimeo. Morozov, I and E. Podkolzina (2013) Collusion Detection in Procurement Auctions, Basic Research at the National Research University Higher School of Economics (HSE) WP BRP 25/EC/2013 The European Commission Green Paper, 1996. Public Procurement in the European Union: Exploring the Way Forward. http://europa.eu/documents/comm/green_papers/pdf/com-96-583_en.pdf. Racca, G M, R. Cavallo Perin and G. L. Albano (2011) “Competition in the Execution of Public Procurement” Public Contract Law Journal 41, 89-99. Sorin, S. (1995) “A Note on Repeated Extensive Games” Games and Economic Behavior 9, 116-123. Wen, Q. (2002) “A Folk Theorem for Repeated Sequential Games” Review of Economic Studies 69, 493-512. Wen, Q. (1994) “The Folk Theorem for Repeated Games with Complete Information” Econometrica 62, 949-954.
11
